A Mathematical Model of Wire Feeding Mechanisms in GMAW Investigations into both aluminum and steel welding wire focus on elements that contribute to poor feedability BY T. M. PADILLA, T. P. QUINN, D. R. MUNOZ, AND R. A. L. RORRER, ABSTRACT. A Hertzian-based contact model ofthe friction between the welding wire and the wire liner has been developed to predict the wire pulling force for gus metal are welding (GMAW). The model predicts a 2.5:1 exponential increase in wire pulling force as the bend angle be tween the welding wire and the wire liner is varied from 0 to 180 deg. A one-dimen: sional force transducer recorded the wire pulling force, while a linear actuator pulled the welding wire through various wire liners at constant wire feed speeds ranging from 10 to 18 mimin (400 to 700 inJmin), Results indicate model agree: ment to within + 0.60 N (0.14 Ibt) RMS ‘when using aluminum welding wires rang: ing from0.8to 1.6mm (0,030t00.0625 in.) indiameter. Extension of the model to ap plications feeding steel welding wires ex hibits agreement to within + 0.80 N (0.18 Ibs) RMS when using 1.1-mn-diameter (0.045-in.) welding wires and a spiral ‘wound steel wire liner. Statistical analyses show an independence of wire pulling force from wire feed speed over the inter val of speeds ranging from 10 to 18 mymin (400 to 700 in min). Introduction ‘A key factor in the performance of gas meal are welding (GMAW) is the feed: ability ofthe welding wire through the Wire liner. For aluminum electrodes, fu tuations in the wice feed speed (WES) of even 1% can lead to iregular arc lengths, oscillatory voltage and current levels, and degradation of the overall weld quality TM PADILLAandD.R. MUNOZ are with Co. ‘orado Schoo! of Mines, Dion of Engineering Golden, Colo. P QUINN is with National In stint of Standands and Teno Materials Re- libiiy Disision, Boulder Colo RA L- ROR TRER ts with Univesiy of Colorado at Denver, Collegeof Engineering, Denver, Colo. Contribution of NEST, an agency ofthe US. Government not subjectiocopyight inthe Unied States. FE 42008 (Ref. 1). In part, variations in the target WES are caused by adverse conditions uring welding (Refs. 1), These include damaging effects such as “stickslip” mo- tion of the welding wize, premature wear ff the contact tube and, in general, un- controlled movement of the hose package uringwelding. A considerable amount of elfort (Refs. 5-9) has been spent to de- velop microprocessor-based wire delivery systems and specialized components that are capable of reducing velocity variations to within 0.5 to 1.0% of the target value. Here, these specialized components and processes allow for controlled feeding at stances of up co 60 m (200 ft) between the wire spool and the workpiece. How- ever, the use of such systems is an expen= sive and nontrivial option when compared o conventional means of wire delivery (Ref, 10), thereby leaving eoom for more practical and cost-effective solutions. ‘At present, a limited amount of pub- lished research has been conducted to un~ derstand and separate the various ele- ‘ments that contribute to poor wire feeding performance. One important factor that controls wire feedabilty is friction be- tween the welding wire and the wire liner. Large or irregular wire feeding forces have been shown (Refs. 3 and 4) to cause pronounced fluctuations in the WFS and ‘wire buckling. These problems are more readily manifested when feeding ahi- minum and other difficut-to-feed alloys KEY WORDS ‘Aluminum Wire Feeding Friction Model GMAW Wire Feeding Hertaian Contact, Mechanics Stee! Wire Feeding Wire Feeding Force Wire Friction (Ref, 11), Studies have been conducted (Refs, Gand 12) to measure the wire feed- ing force during welding, but a limited ‘amount of detail has been given to model the source of friction, This paper presents a Hertzian-based contact model of the ‘welding wire-to-wire liner friction as the welding wire is pulled through various wire liners at constant velocity. Model Movement ofthe welding wire through the wire liner is similar to "belt-on-drum” applications — Fig. 1 Historically, thissit- vation has been ‘seated using Euler's equation (Ref. 13) for rope friction given by T= Tel? w where 7; refers to the slaci-side tension, T, the tight-side tension, . the wire-to- liner eoetticient of friction, and the wire- to-iner contact angle. Predictions of the wite-t-liner fiction force using Euler’s equation rely on the use of known values for the coefficient of friction. Even when the formulations of the welding wire and wire liner materials are known (most are proprietary), little data exists for the coefficient of friction in the context of a wie sliding on a station- ary surface, Predictions can be made using the coefficient of friction from similar mae terials, but these predictions do not ade «quately fit the experimental results and ei- ther overestimate or underestimate the ‘ruc functional form ofthe data (Ret. 14) ‘An alternative model utilizing aspects of Hertzian contael mechanies and frition theory has been developed to better pre- dict the wire-to-liner friction using only basic information about the welding wire land the wire liner. The alternative model is based upon experimentally determining the shear strength at the welding wire and Wire liner interface, Asa result, there is no need to rely on tabulated dat for the co-a Teovion Ty Fig. 1 — Schematic of Euler's rope friction equation for “belt-on drum” ap pleatons efficient of friction since an actual inter: face parameter can be computed for any welding wite and wire liner pair. The fol- lowing details this development. “According to Hertzian contact theory, ‘when two rough surfaces come in contact with each other, the surfaces distor locally tand make contact at discrete points (Ref 15). Inthe ease ofthe welding wite and the wire liner, the combination can be mod- cled as a pair of axially aligned cylinders that make contact under constant curva ture R, and witha specified “wrap” or con. tact angle 0 — Fig. 2 Ifa differential element of the welding wire is pulled with applied tensions 7) and T (Fig. 3) at steady-state conditions (with assumed constant velocities and negligible accelerations) over constant curvature R and through a fixed contact angle 0, the distributed load w acting per unit length lover the segment of welding wire ean be formulated from a static foree balance Considering the arbitrary orientation of a differential segment of the welding wire (Fig. 3), the equations of equilibrium can be written as Dhea 0; Tan) -ntdoon{ on 48) spied? rear) slo as)=0 o Fig, 2— The welding wie and wite liner modeled as a pir of aly aligned ‘liners under constant curvature and witha specified contact angle ‘Fig. 3—A segment ofthe welding wire and the wire liner showing the contact area caused by the ap- pied tensions Ty and T LA =6 Fo) nts oe odo rar) e) ra. evte-(rsarje-o. Here, the angle @ refers to the initial start position of curvature as measured from tho abscissa, 9 refers to subtended angle of the differential clement, T the applied ‘welding wire tension, 47 the additional WELDING JOURNAL BEERSWreLine Nweappat ata mace Fig. ¢— Experimental apparatus used to measure the pulling force as the welding wit slides throug the wire ner tension due to the tangential resistive force FR the constant bend radius of cur. vvature, and w the distributed loading act ing over the differential element of the welding wire. By small angle theory (Ref. 16) asthe value of d0 approaches zero, the sine and ‘cosine terms containing the a8 argument behave as “tim solts}=ao iol Using Equation 5 with Equations 2 and simplifying to include only first order ferms results in equilibrium equations of the form waco} 1 sso “ ~r40, and the interfacial shear strength 5, FEE 42008 2) Here, the quantities 7), R, Oj, 03,R’, and E* can either be obtained diectly or com- puted from the initial wireto-liner setup, thereby leaving + as an unknown parame= {er that can be determined empirically. ‘The end resul i an alternative model (0 the Euler rope friction equation that is in- dependent of the nominal coefficient of friction and that is capable of predicting the wire pulling force given only basie n= {formation about the welding wire and wire liner system. Experiments Full-scale tests were conducted to de- termine the validity ofthe contact model developed in this research, The welding wire and the wire liner were wrapped around a variable-angle jg (Fig 4) to sim- tlate a combination of bends typically found in the wire-feeding hose package. One end of the welding wire was attached {oapulley and masssystem to create a uni- form backtension on the welding wire, thereby simulating the drag normally as- sociated withthe wire spool. The uniform backtension also helps ensure complete contact of the welding wire with the wine liner. “Travel speed of the welding wire ‘through the wire liner is controlled using & computer-driven linear actuator that is ca- pable of providing a stroke of 3 m (10 ft) A velocities ranging from 0.01 to 30.5 ‘mimin (05 to 1200 in,/min) to within = 5% absolute. Once a particular combi- nation of the welding wie, wire liner, wire feed speed, and bend curvature was se- lected, the linear actuator was initiated aand triggered a data-uequisition system to record the wire pulling force. The force transducer used in this application is ea- pable of measuring up t0 110 N (25 Ibf) Wwith an uncertainty of = 005% full-scale as computed from astaic calibration, Fol- lowing a force measurement, the wire jig was reset to a new contact angle and the ‘experiment was repeated. Tn this study, three individual cases were considered to determine the aceu- racy ofthe contact model in predicting the wire pulling force when using various com- binations of readily available welding wires and wire liners. In each ease stud, the welding wire was initially unspooled With a wire east! radius of approximately 50 mm (2 in.) and subsequently straight- fened to a cast radius of approximately 1.2 1m (48 in.) Helix! in the welding wire was ‘measured to be less than 13 mm (0.5 in.) aand was considered to be negligible, The surface of the welding wire and the wire liner was dry, unlubricated, and used in the as-manufactured condition, Replica- tion measurements (shown as data error boars in attached figures) were laken for statistical analysis at wire-to-liner contact angles equal t0 0, 60, 120, 140, 160, and 180 deg, respectively. Each case is out lined as follows and is summarized in Table I for larity Case No, 1 —Matched-Size Aluminum Welding Wire Dry-Sliding on a Tetlon® Wire Liner ‘An aluminum welding wire was used with a matched size (as recommended by the manufacturer) polymer-based ‘Teflon? impregnated wie liner. Experiments were 1. °Case refers tothe amount of natural clin the velding wire de othe wrap of the wire oma the Spook “Heli” refers tothe amount of natural oil Iihe welding wire caused by He ltral wrap of the wire ram one sie of the dram to anather, 2 Trade names contained herein are used only 0 generically identify the material used in ts re Search Actual welding wire and wie liner compo- Stions are proprietary and hence unknown. Such identification does not constiute recommenda tion or endorsement of these materialsa toa a Senge BP ‘We Tees pees Fig. 9— Comparison ofthe interacial shear sength values aa function of WES for @ 18:mm (0.030.in,) ERS356 sold-core aluminum welding wire drying oma 137 (0.052-n }imeral diameter Teflon wire lner Fig. 10-— Experimenta and theoretical comparison using pooled data resus, for a 08:mim (0030+) ERSI96 sold-core aluninurm welding wie dy sliding on a 132mm (0082) intemal diameter wie liner Uncertainty the model is own as dashed lines and represents ne standard deviation in the estimate ofthe interfacial shear strength. The data shown have been nor ‘malied fra 5.N (11-1) uniform backtenson fo show anonzer force value at zero contact angle representing static friction i aa i [7 : ace tite i tee ' it Hf [pa a ie iF ete i Hee : fae H i if ele . “= ly . 7 ma ares Cones 1g 1 Experimenta and hear comparion wing Ln (0.0825. Fi 12— We paling fre fr WPS = 101 18 min (4002 700i) in) BRS3S6 solid-corealuminurs welding wire drysiding on a 2S: (©.100-n) internal diameter Nylon wie liner. Uncertainty in the model is Shown as dashed lines and represents one standard deviation nthe estimate ofthe interacial shear strength. The data shown have been normalized for a SN (1-29) uniform backenson ta show a nonser force vale at zero con ‘act angle representing static fiction witha T.l-mm (0045-n,) ESTOS soidcorecopper-lad ste welding wie dhy-slding ona 1.7mm (0.065-n) intemal diameter spiral wound sel wie liner The daa shown have been normalized for @ 10.N (226f) uniform backtension to show a non-sero force vale ro contact angle representing sate friction WELDING JOURNAL BEERS|| | ‘Win Feed sped mi Widng eae Fig 13 — Comparison ofthe interfacial shear srengh values as a function OF WES fora LJ-mm (0045: ERTOS6 sold-corecopperclad tel weld ing ite dy-slding on 172mm (0.065) intemal diameter spiral sound see wire liner ‘ig. 14— Experimental and theoretical comparison using pooed data results fora Lm (0045 in) ER7OS6 ole core copper clad sel welding wire dy lng or 1.7m (2.065. itera diameter spiral wound steel wire ner ‘Uncertainty inthe mode is shown as dashed lines that epreten one standard deviation nthe esonate ofthe interfacial shear sength. The data shown have Deer nomad fora 10-N (2.2406) uniform backtension to show @ nonzero force value at ero contact ange representing static fiction. conducted using a 0.8-mm (0.030-in.) ERS356 solid-core aluminum welding wire spooled onto a standard L-kg (2-Ibm) drum, A 25 m (8 ft) length of 1.3-mm (0.052-in.) internal diameter “Teflon”? wite liner was used and had the manufac. turer’ rating for use with welding wires ranging fom 0.8 to 0.9 mm (0.030 t0 0.035, in.) in diameter. This particular combina tion of welding wire and wire liner pro vided a diameter ratio of 1.7: Wire feed speeds were varied from 10 to 18 m/min (400 to 700 inJmin) in 2.5 m/min (100 in/min) increments to repre sent the middle range of wire feed rates used during typical aluminum GMAW. Bends in the hose package were simulated by varying the contact angle between the welding wire and the wire liner from 0 to 180 deg in 10-deg increments. A S-N (1.1 Ibf) uniform backtension was applied to ‘the welding wire withthe pulley and mass setup to ensure complete contact of the welding sire with the wire liner and to Simulate drag normally associated with the wire spool 43. “Telon “and "Nylon been sed asa shor hhand notation to refer to the polymer-based, Teflowimpregnated wire liners and the polymer. based, Nsom impregnated wire mer sed in is studs | 106-s Ion Cate No, 2— Matched-Size Aluminum Welding Wire Electrode Dry-Sliding on a Nylon Wire Liner Case No. 2tested the applicability of the contact model to a larger diameter (= 2X larger diameter than that for Case No. 1) aluminum welding wire and polymer-based, Nylon-impregnated wire liner. This study ‘was also conducted to determine whether there was an improvement or degradation in the overall wire feeding performance ‘when using the "Nylon"® wie liner. [Experiments were conducted using a 1.6-mm (0.0625-in.) ERS356 solid-core aluminum welding wire spooled onto a standard 1-kg (2-Ib) spool. The welding wire was guided through a 2.5 m (8 ft) length of 2.5-mm (0.100-in.) internal di- ameter Nylon wire liner. This particular Size of wire lines was recommended exclu sively for use with 1.6mm-diameter weld- ing wires and provided a diameter ratio of 1.6:1. An intermediate constant-valued WES of 14 m/min (550 in.min) was se~ lected, This single constant-valued wire feed speed was selected to give a uniform metric of comparison in determining ‘whether there was an increase or decrease in wire pulling force performance when using the Nylon wire liner. The contact angle between the welding wire and the Wire liner was simiterly varied from 0 to 180 deg in 10-dleg increments. A S-N (1.1 Ibf) uniform back-tension was again ap- plied tothe welding wire. Case No. 3— Matched-Size Copper-Clad Steel Welding Wire Dry-Sliding on a Steel Wire Liner Case No.3 was conducted to determine the applicability of the contact model to steel-on-stee] GMAW applications when feeding stel welding wires of intermediate diameter through a spiral-wound steel wire liner. Experiments were performed using a 1.1-mm(0.045-in.) ERT0S-6 solid-core eop- per-clad steel welding wire spooled onto a standard 05-kg (I-lb) drum. A 2.5m (8 ft) Jength of 1.7-mm (0.065-in) internal diam ler spiral-wound steel wire liner was used, This particular liner was recommended for use with welding wires ranging from 0.9 to 1.1mm (0.035 to 0.045 in.) in diameter and provided a diameter ratio of 1:1. Wire {eed speeds were similarly varied from 10t0 18 m/min (400 to 700 in.min) in 25-mm'min (100-in./min) inerements, while bends in the hose package were again varied from 0 to 180 deg in 10-deg increments. The wni- form backtension on the welding wire was increased to 10 N (22 Ib) to compensate forthe relatively stilfer steel welding wie,etoecioet EPS CRG _ AN Fig 15 — Speculaeddifence inthe nature of contact fora sel stifer) velding wire vs. an alumina (more fle) welding wire Fig 16— Hypothetical mixed asperiy-o-asperiy contact of the welding wie th he wireline showing potential merfocal shear zones, Results and cus nm Case Study No. 1 — Matched Size Aluminum Welding Wire Dry-Siiding on a Totton Wire Liner The wire pulling force data (Fig. $) in dicate there is an approximate 2.5:1 expo: rental increase in Wire pulling force asthe contact angle between the welding wire and the wire liner is increased from 0 to 180 deg, The severity of the bend in the hose package (ic, an inerease in the an gular sum of bend) increases the mean Wire pulling force. This corroborates pre- vious research (Ref. 12) that the configu: ration of the hose package plays a signifi ‘cant role in wire feedability. Normalizing the data set with respect (o the applied Dbucktension, the nonzero force value ob- served at zero contact angle is believed to represent the minimum static friction in the wire-to-iner system. Its believed the Stati frition component of force is due to sell-weight of the welding wire acting against the wite liner and, more predomi nately, the entry/exit effects ofthe welding wire passing through the wire liner. If the data are directly compared against Euler's equation for rope friction (Equation 1) using a representative range ‘of coefficient of friction values (= 0.210 1 = 044), the measured wire pulling force ‘can be greatly underestimated or overesti- mated — Fig. 6. Here, the representative coefficient of friction values are typical (Refs. 17, 18,20) and include pairs similar to the tested wire tnd liner system, such as polymer-on: Steel, aluminum-on-Teflon,aluminum-on- Nylon, and Plexiglass-on-steel. A problem ‘with using typical cocificient of frition data is compounded by the wide range of coefficient of friction values that depend ‘on whether the system is lubricated or un- lubricated. Asa result, it isnot clear which coefficient of friction value adequately represents the true welding wire and wire liner system under consideration, Inorder to directly compare the exper imental results with the contact model, ‘while avoiding the coefficient of friction ambiguity, the “Bootstrap” method (Ret. 21) was used to determine the specific in- {erfacial shear strength (ISS) + for the tested aluminum-on-Teflon system. In brief, the Bootstrap algorithm isan -iter- ation method of determining the average value and uncertainty of an unknown model parameter (see flowchart, Fig. 7). The algorithm uses an initial least-squares estimate of the unknown parameter, fol- lowed by iteration to generate asecondary guess of the unknown parameter. The sec- ‘ondary guess is based upom a faux data set that has randomly been assigned a resid- ual error computed from the first run, The process is repeated m times to give esti- mates of the unknown parameter, from ‘which a sample average and standard de- Viation ean be computed — Fig. 8. This provides a measurable way of determining the uncertainty in the model. In this case study, 5000 iterations of the Bootstrap al- gorithm were performed to determine t fand ils uncertainty at each of the four Wwire-feed speeds — Fig. 9, As indicated in Fig. 9, the value + is grouped around a central value of 0.9 MPa (130 Ihjin2), prompting a mull hypothesis to suggest the interfacial shear strength values (each for a different WFS) origi- nate from the same population, thereby indicating an independence ‘of wire pulling force from wire feed speed. A Chi- Squared goodness-of-fit test was con- ducted with the interfacial shear strength values using 4 95% confidence interval (v = 3, a= 0.05) toyield a computed statis- tic of 0.262 compared to a critical statistic of 5991, ‘Accordingly, the null ypothesisis con- sidered true and indicates the wie pulling force is independent of wire feed speed lover the range from 10 to 18 rmin (400 10 700 injmin). As a result, the wire palling force data of Fig. 5 were pooled as fone data set and passed through the Boot- strap algorithm (with n = 5000) to gener- fate a new estimate of the interfacial shear strength equal to 0.9 MPa = 0.1 MPa (130 Tbyin.2) for the composite data set. “The interfacial shear strength estimate was subsequently used to compare the pooled data set with the contact mode! — Fig. 10, Here, the solid line indicates the ‘mean wire pulling force (meun estimate of +) and the dashed lines indicate the un- certainty in the model based upon the un- certainty in the estimate of the interfacial shear strength. As shown, the contact ‘model is within = 0.60 N (0.14 bf) RMS ff the measured values and matches the 25:1 exponential increase in wire pulling force fora contact angles ranging from 0 10 180 deg, Case No, 2—Matched-Size Aluminum Welding Wire Dry-Siiding on a Nylon Wire Liner Since the results forthe aluminum-on- ‘Teflon study indicated the wire pulling WELDING JOURNAL BEESforce was independent of wire feed speed, wwe assumed the same conditions would hold true when testing aluminum-on: Nylon. As a result, a single constant wire feed speed of 14 m/min (550 in min) was selected for this study to give @ uniform metricof comparison of wire pulling force performance when using the Nylon wire liner. “To compace the values ofthe measured wire pulling force, the aluminum-on Nylon data were similarly passed through the Bootstrap algorithm to generate a ‘mean interfacial shear strength of 0.5 MPa + 0.1 MPa (73 lbfin215 Ibn.) Substi- tuting this computed value of + into the ‘contact model results ina similar 27:1 ex ponential increase in pulling force when tsing the 2X larger-diameter welding wire tnd Nylon wie liner — Fig. 11. The com. parison again indicates the severity of the bend (ie, an increase in the total angular sum af bends) inthe hose package plays a significant role toward increasing the ‘overall wire pulling fee. In general, the model predicts the wite pulling force to within = 1.31 N (0,30 Ibf) RMS of the measured value. Likewise, the nonzero force component at zero contact angle is again believed to represent the state frie tion in the wire-to-liner system. In this ‘ease, the static frietion contribution is towice as large as that ofthe aluminum-on: Teflon system, When predicting the wire-to-linr fric tion, the contact model provides abreadth of distribution that is 0% greater in pe dicting the wite pulling force when using the welding wire of larger diameter (Fig, 10vs. Fig. 11), The cause of the larger var ation is ot known, Despite the inerease in uncertainty, the predicted and measured force values for either ofthe two wite lin. ers (Nylon or ‘Teflon) lie within 20% of each other, with the Nylon wire liner pro: ducing the overall larger sire pulling forces for similar comtact angles. Case #3 — Matched Size Copper-Clad Steel Welding Wire Dry-Sliding on a Stee! Wire Liner Since many of the same aluminum GMAW wire feeding problems are expe: rienced in steel GMAW, Case No. 3 was ‘conducted to assess the contact model in predicting the wire pulling force when using a copper-clad steel welding wire and a spiral-wound steel wire liner. The wire pulling force for a copper-lad steel wire Gry-sliding on a steel wire liner is again seen (Fig. 12) to increase as a function of ‘contact angle, but does so nearly lineatly (R2=0,96) at arate of 2:1. Again, the data indicate that tighter bends in the hose package increase the wire pulling fore. ‘The Chi-Squared goodness-ol-fit test was repeated with the stecl-on-steel data FEE 2008 sets to again show mathematically (with 98% confidence) that the computed val- tues of interfacial shear strength for each wire feed speed (Fig. 13) originate from the same population as evidenced by a computed Chi-Squared statistic of 3.891 ¥. a ctitcal statistic of 5.991, As a result, the wire pulling force forstecl-on-steel ap plications is aso shown to be mathemati- cally independent of wire feed speed over the interval of speeds ranging from 10 to 18 m/min (408 10 700 in.min), despite a 10% increase from one wire feed speed 10 another. Using the pooled data set for Seel-on-steel, the mean interfacial shear strength was computed to be 3.9 MPa = (0.4 MPs (566 Ibjin+58 tbjin2) The wire pulling foree data were simi larly pooled and directly compared with the contact model predictions — Fig, 1, As indicated, the contact model predicts ‘an inerease in wire pulling force with in- creasing contact angle and again suggests that tighter bends in the hose package in- crease the mean wire pulling force. Like wise, the static friction ia the wiee and liner system is observed to be similar to that of the aluminum welding wire sliding through the Nylon wire liner. The increase in wire pulling force inthe case of stecl-on-steel is observed to in- erease more linearly than was previously exhibited during the test of aluminum-on- Teflon and aluminum-on-Nylon. Despite this difference, the measured and pre~ dicted quantities are within + 0.80N (0.18 Ibf) RMS of each other. In general, the steel-on-seel results produce mean wire palling forces that are up to 10% larger than that for either of the two previous aluminum tests ‘We speculate that the difference inthe observed behavior of the steel and ali- ‘minum can be attributed to the nature of the contact between the welding wire and the wire liner. Since Young's modulus of the steel welding wire [Ey = 20) GPa (30 Mpsi)] is approximately three times that of the aluminum welding wire [E,) = 73 GPa (10 Mpsi)}, under identical applied tensions of T; and 7, welding wires of nominally identical diameter may fail to make complete contactor only make par- tial contact withthe wire liner — Fig, 15 Asa result the loading distribution devel- ‘opment (Equation 10) and the resulting governing differential equation (Equation 21) would undoubtedly difer for steel and aluminum conditions, and may account for the observed exponential and linear sifference in the data, We also speculate that the interface be- tween the welding wire and the wire liner plays a significant rote in determining the ‘wire pulling force. In all case studies, the computed interfacial shear strength is lower than theoretically expected, Micro- scopically the interface between the weld- ing wire and the wire liner can be consid- cred to consist of mixed asperity-o-asperity contacts — Fig. 16. Asa result, one of sev- eral or combination of potential shear zones define the true interfacial shear strength ‘Although the contact model is devel~ ‘oped to account for an “average” interfa- cial shear strength value over the entire contact region, the theoretical basis ofthis ‘model only accounts for “bulk” values of the welding wire and the wire liner. Con- sequently, the contact model does not se- count for additional intricacies con tributed by interactions caused by the oxide layer (if present, lubricant (if pre- sent), oF any other higher-order factors such as “composite” material values or third-body debris, Conclusions 1) The welding wire and the wire liner can be modeled as two axially aligned ojlinders utilizing aspects of Hertzian con- ‘tact mechanics and frition theory to com ppute the interfacial shear strength at the ‘wirelner interface. The contact model in- corporates the bend angle and radius of curvature between the welding wire and the wire liner, as well as nominal diame- ters and moduli of the welding wire and wire liner. 2) The contact model accurately pre- dicts the wire pulling force for aluminum and steel GMAW applications to within + 0160 N (0.14 Ibf) RMS and + 0.80 N (0.18 lof) RMS, respectively. A statistical analysis has shown with 95% confidence land regardless of liner composition, the Wwire-pulling force for aluminum-on- ‘Teflon and steel-on-steel GMAW applica- tions is independent of wire feed speed ‘over the interval of speeds ranging from 10 {0 18 m/min (400 to.700 in min) 3) Inall tests, the wire-o-ner friction force increases as the severity of the bend in the hose package is increased. This in- dicates tighter bends in the hose package require larger wire pulling forees to over- ‘come the inereased friction, Additionally, the nonzero force component at zero con tact angle represents the static friction in the welding wire and wire liner system, When testing aluminum-on-Teflon and aluminum-on-Nylon wirelliner combina- tion, the wire pulling force increased ex- ponentially at a rate of 25:1 and 27:1, re- spectively. A 20% larger difference was lobserved between the mean wire pulling force values when using Nylon-impreg= nated vs. Telon-impregnated wire liners AAs well, under similar conditions, the wire palling forces for stee-on-steel wireiner combination were observed to increase more linearly at arate of 2:1, and in gen= eral were up to 10% larger than either combination of aluminum welding wieesAckoosedements Journal 4(2) 1185 to 121s Academie des Sciences de Bern. pp. 265-278. S:Sumokovlishy,D, A.1986,Wiefeedsys 14, Padilla, . M. 2001. A mathematical Theauthorsthankthe American Weld- tems forrobotic MIG welding. Metal Conszuc- model of wire feeding mechanisms in gas metal ing Society's Graduate Research Fellow. om 18(5) 293-296 ace welding, MS, Thesis. Colorado School of '. Nichols Research Corp (NRC).1992.De-_ Mines, Goldea, Coto. ship Program and the NIST Professional Recenek Expense Programs (PREP) ‘meat ofan aomsted wire dlvery yee TS. Greenwood, JA and Wiliamson,1.B tem for robotic welding applications, NASA- P1996, The contact of nominally at surfaces. for project funding. ‘The authors also ex Re394216 37, Proceedings ofthe Reyal Sociey: London, Vo. tend their most sincere thanks to Pr. Tom 7, Kuvin, BF 198. New wite feeders de~ 295, Series A! 300-319 Siewert of NIST and to Mr.John Hinrichs ver the goods. Welding Design and Fabrication 16. Edwards, C. 11, and Penney, D.E, 1990 of Friction Stir Link, Inc, for their practi-Ti(2):29-82 CCaleatus and Anabieal Geometn, td ed New ‘al insight and advice into this research 8 Villauerte, J. 1999. Understanding con- Jersey: Prentice Hal, ne. tact tip longevity for gas metal are welding 17, Willams J. , 1996, gineering Tibol eling Journal 712): 29-38 ‘og. New York: Oxford University Pres Inc References 9, Alms, D, 2000, Wie-leed upgrade drives 18. Bowden, FP, and Tabor, D, 1980. The sluminum-welding productivity. Metal Forming Picton and Ladrication of Solis, Vl. 1-Oxford 1. Quinn, 1: P. 2002, Process sensitivity of 34 (8) 28-32. Pres. GMAW: aluminum versus steel, Belding Jour 0, Hinrichs, 1 F198, What's wrong with 19, Chills, TH. C. 1980, The contact and nal 81(8): 55-10 64s. sas metal are welding (GMAW)? Dealt Revie friction between flat belts and pulley. ntera= 2, Rods, JF, Brown, D.C, and Groth, W, Sion The Welding Link ional Journal of Mechanical Science 22 G. 1966. Stuy of current contct tubes for gas Hinrichs, J. F, Noruk,J.S..MeDonald, 117-126, metal are welding, Welding Journal 45(8): 378 W.M,,andHleideman, R.J.1998,Challengesof 20. Marks, L.S, 1916, Standard Handbook toate ‘welding aluminum alloys fr automotive struc- for Mechanical Engineers. New York, NV Mc- 3, Yamada T and Tanaka, ©, 1987, Flac tures. Svtsare 503) 7-9, GrawHil ation ofthe wite fed rate in gas metal arcwel 12, Dithey, U. and Reisgen, U. 1995. Tests 21. Bradley, F. 1982, The Jack Kaife, the ing. Welding Jural 66(9): 342. lon wie feed jstems for gasmeialare welding. Bootstrap and other resampling plans. Society 's-Quinn, TP, Madigan, RB, Morn, M, Welding and Cuting 47(): B12-E14 for Industrial and Applied. Mathematics, ‘A. and Siewer, A. 1995, Contact tube wear 3 Euler, L. 1762. RemarguessurVetfet du. CBMS-NSP Regional Conference Series in Ap delection in gas metal are welding, Welding frottement dans Tequilibre. Memoires de pled Mathematics. Philadelphia, Pa, 38 p. SS WELDING JOURNAL BEERS